20-300Hz: Eminence Alpha 15"
300-1500Hz: Vifa P13WH
1500-20kHz: tbd (Vifa DX25, XT25, or DX27)
Actively crossed and tri-amped.
Follow this design tread at DIYAudio.
This design had reached a plateau of performance when trying to equalize the midrange from 300Hz and above. Large baffle caused undesirable peak and dip which resulted in rough midrange. Baffle size should be no more than 2x cone diameter for smooth response. Refer to S7 for the final design to fix this issue.
Eminence Alpha15A crossed 24db at 1kHz. Good tonality was achieved but both drivers especially the vurnerable TC20 are strained.
What I gained from these prototypes are 1) Excellent bass using the Eminence with correct shelving lowpass filters, 2) Eminence peaks and their appropriate notches, 3) The Alphas have surprisingly good upper frequency extension.
This measurement was taken from listening position, in-room. Ignore the SPL numbers as I did not calibrate them.
sudo add-apt-repository ppa:nilarimogard/webupd8
sudo apt-get update
sudo apt-get install audacious
and if encountering error:
rm -r ~/.config/audacious
For slow computers like mine (PIII running Ubuntu Lucid via USB 1 !!):
- Choose Output-Only (not Duplex) from Ubuntu Preference->Sound
- Select PulseAudio, 10000 buffer, and 16 bits from Audacious.
My music collection has been growing, and in a good way. About 250GB of which 70% are native 16/96kHz and 24/96kHz recording in lossless FLAC format. Mostly sourced from Vinyl/LP rips. They are superior than normal CDs which are recorded at 16/44.1kHz.
As on-the-shelf audiophile media center or server is very expensive (read: Linn DS is $15,000!) I want to have a media center to allow me sit back and enjoy the music without fiddling with laptop keyboard as now.
I tried many Linux implementation of media center and finally settled on MyMediaSystem. It is pretty, fast, and useful with little non-needed gimmicks. There are additional 'bonus' such as movie player, TV player, internet radio, and photo viewer.
This is what I did. The hardware are:
IBM Thinkpad T23
DAC: SoundBlaster 24 bit Live!
Ubuntu 10.04 or above
MMS installation how-to
- Install Ubuntu then run updates.
- Install Ubuntu Restricted extra (to be able to play mp3 etc.)
- Add the following lines from Accessories > Terminal
sudo add-apt-repository ppa:mms-prodeia
sudo apt-get update
sudo apt-get install mms
sudo apt-get install mms-plugin-audio
- Edit the AudioConfig file in /etc/mms/. From Terminal type:
sudo gedit /etc/mms/AudioConfig
edit this line (after mapping a drive as explained here)
audio_dir_no_watch = /media/wdtv/04. MEDIA/01. MUSIC
audio_player = AlsaPlayer
If the collection is on network drive like the example above, one needs to use smbmount as in this article here or here
Note: The SoundBlaster Live needs to be selected as default from System >Preference > Sound. In the past, this was done using asoundconf (retired now)
The above option installs a minimalist version of the Media Center basically audio player and Internet radio.
To share directory
Install smbfs: sudo apt-get install smbfs
Create mydriveconnect.sh and make it executable
sudo mkdir /media/mydrive
# unmount previously-mounted drive
sudo umount /media/mydrive
sudo smbmount "//192.168.1.11/wdtv_data/04. MEDIA/01. MUSIC" /media/mydrive -o username=defaults,password=defaults
#sudo smbmount "//192.168.1.10/sh_widhibrata/_backup/04. MEDIA/01. MUSIC" /media/mydrive -o username=widhibrata,password=password
in case RW is desired, use uid and gid option:
sudo smbmount "//192.168.1.11/wdtv_data" /media/mydrive -o username=default,password=default,uid=999,gid=999
find out the uid using "id" command from terminal
With Lubuntu, install xfce4mixer to select sound card:
sudo apt-get install xfce4-mixer gstreamer0.10-alsa
Slightly more heavyweight - you could install pulseaudio and thus use the pulseaudio packages such as
sudo apt-get install pulseaudio pavucontrol
- Coherent sounding due to no crossovers
- Somewhat acceptably clean due to low distortion above 70Hz
- Midrange, vocal is warm and glorious. Very realistic
- Small size
- Easy to make
- Low efficiency
- Box signature unavoidable
- Box quality is poor
- High-frequency beaming
- High frequency not sounding 'sublime'
It's ok... if I really want to serious playback I go to my living room for the cannons.
I use Dick Smith 5" monitor bought at ebay auction for $10. With the Tangband bamboo driver it was a lucky match! Ignore the magnitude, it's not a 100db efficiency but observe the driver rolloff which is then extended to about 45Hz by the port tuning. In fact this loudspeaker low extension is fine and I don't feel the need for a subwoofer. In my 3x4m room they are loud enough. The TangBand is flat up to a dip at 15khz but my measurement was not ideal as it was taken in front of the grill and there are diffraction effects. Also the drivers are not flush-mounted (yet).
Harmonics distortion measurement at 50Hz shows reasonable H2 and H3 at 100 and 150Hz. Ignore that peak at 5kHz as it is my computer fan noise being picked up by the microphone. First audible signal is heard around 40Hz. There is no audible sound at 20Hz (as it should) and 30Hz.
More comprehensive distortion measurement. At around 70Hz the 2nd and 3rd harmonic distortion drops to about 1% level which explains the average cleanliness of the sound. Most of the time H2 dominates which is normal cellulose cone behaviour and gives somewhat warm sound. H2 or 2nd harmonics is more pleasant than 3rd. I drive them with little Dared MP5 amps.
Photo of the dicksmith box. It is not well-braced and that's next on my to-do list. The measurement is:
(outside measurement, panel thickness is 12mm)
Port diameter: 50mm
Port length: 110mm, flared on the outside
Closeup of the TangBand W3-1364 3" speaker. The cone is made from bamboo fiber and the chassis is cast metal (!). Very well made. This driver is highly regarded albeit the cost. See full review at ZaphAudio and his measurement results.
The measurement lab. As usual everything here are prototypes, temporary, and utilise anything available in the garden shed :)
- Added Baffle Step compensation (BSC): 1mH parallel with 8.2R and connected in series with the speaker. Theory and design is here.
- Few attempts to add more damping in the box resulted degradation in sound quality. Better left alone and live with the colorations.
- Blocked the port to make them sealed-box speakers. The bass sounds more realistic, albeit less extension.
- I spotted a very nice build of similiar small TangBand family here: http://attlid.eu/p_swift.html
Never embark on a loudspeaker project without decent measuring tool. This is my set of tools to measure speaker performance. My criteria is simplicity and accuracy.
- Sound Card
Soundblaster Live! 24 bit - $60
ARTA. The demo is free and fully functional, except you cannot save files
Alternatives: HOLMImpulse, REW (Room EQ Wizard), Speaker Workshop. All free.
DIY Panasonic WM-61A microphone - US$3 (plenty from ebay)
Alternative: Behringer ECM8000
To utilise the WM-61A is quite easy. Electret mics all needs "phantom power" and the circuit is something like this, just need a 9-volt battery, 2k2 Resistor, and 10uF Capacitor (non polar). Of course if you want the real deal you can create something like SL's measurement mic or something like this.
The output plugs to left channel if USB DAC's "Line-in". On stereo 35mm connector it's the pointy bit. Mic's ground is the one connected to the body.
The finished product! ... I use chopstick to make it more tasty.
Close-up pic of the WM-61A. They're only like $3 a piece but very flat!
Download and install ARTA from this website, it's quite easy to use. For more functionality such as enclosure design etc. You need to use Speaker Workshop which is free, but very difficult to setup. I actually tried and gave up.
What I like to measure using ARTA are:
1. Frequency response. Simply point the mic to the speaker, choose "FR1" and click the play button. There will be pink noise heard and graph should appear. Something like this. In this case there were two measurements - the yellow line shows bass reflex port response.
2. Harmonics distortion - When a speaker is fed a tone, let's say 40Hz it should only sound 40Hz right? Well unfortunately no. There will be other tones heard as well and these are called harmonics. Ideally the harmonics should be none. The harmonics came in multiplies so for 40 Hz. ther will be:
2nd harmonics - 80Hz
3rd harmonics - 120Hz
4th harmonics - 160Hz
People say that 2nd harmonics sounds more pleasant than 3rd and above. To measure these harmonics, choose "SA" or spectrum analyzer then configure the generator to sine wave at the required frequency.
3. Harmonics sweep - this is similiar to the above, only that it is performed in sweeps. To do this quit ARTA and in the start menu there will be a program called "STEPS". Similiar to ARTA, run this and you'd get this diagram. In the example below it is clearly shown that the driver being measured performs best above 70Hz, where its harmonics are low. This is TangBand W3-1364 and with distortion of around 1% like that it's quite impressive.
So when all has been measured will the speaker sound good? NO. I view these measurements as a guide only, a debugging tool. Tranducer performance is much more complex than the above measurements.
But those who do not measure, don't know.
I imported quite a few of these mics, and people have been asking for them. I'm always happy to distribute to fellow diyers, but there is none left now. Please do not ask me to send you the mic. And btw. a group buy at diyAudio or SNA may be the way to go (and do let me know if you organise one!)
The microphone can be easily found in ebay:
In 2015 I was asked to commission a pair of loudspeakers for Princeton University 3D3A project. I thought it was a good opportunity to compare my mic's measurement with purpose-built lab.
Fig 1: Backyard measurement using this microphone
Fig 2: 3D3A anaechoic chamber
Fig 3: Sonogram measurement using this mic
Fig 4: Sonogram measurement by 3D3A
Looking at the comparison it is clear that this seemingly simple microphone is up to the task!
- Eminence Alpha 15"
- Vifa Midbass and Tweeters => Replaced with bamboo fiber FR
- XO completed 23/9
- Time alignment completed 23/9
- Notches ...
- new Baffle ...
- PCB ...
Very low extension of dipole bass, reaching 20-30Hz at -3 dB
This is the second iteration of original Stealth Open Baffle speakers. It is practically the same as Stealth OB, aside from additional woofers and obvious baffle size plus a bit better documentation ;). It is a culmination of my work/experimentation with various designs of loudspeakers.
Nothing can compare the naturalness and non-coloration of dipole bass, and the 4 x 10" woofers provide plenty of them. The system acts as dipole up to 6kHz which is considered very high.
How much bass? Well if you have system that goes down low enough and permit my digicam's distortions these clips may give some ideas :)
The baffles are 45 x 60 x 18mm MDF. In Australia go to Mitre 10 and get the 1,200mm one and ask them to cut in half - simple. It cost about $15. I designed them to be placed on top of my study room table. If you want them on the floor simply do a taller baffle. The rule is very simple with OB: The bigger, the better.
I painted them flat-black and it looks good. It's how I got the "Stealth" name from as it does remind me of F117. Yes it also "dissapears" on good recording.
Once you hear dipole or open baffle bass, it is hard to "unlearn" the experience. One would instantly recognise what "box coloration" is and will be irritated by it. Unfortunately dipole bass is hard to produce due to cancellation.
This can be rectified by displacing more air and the way to do it is by adding more woofer, bigger woofers, and woofer selection. The woofer must have Qts value of 0.7 or more (which means less damping, smaller magnet, etc). This tends to be cheaper woofer too! The Stealth 2 uses 4x 10" woofers. They have Qts of 0.7 and cost only $29 from Jaycar.
Second step is to equalize it. The equalising circuit is very simple as shown below. It can be built using three op-amps such as TL072 ($1 a piece). The signal goes through 120Hz low-pass filter and then shelved at 6db/octave to compensate the cancellation. The signal phase is inverted so from the amplifier the cables must be swapped between positive and negative.
It uses regulated PSU using 7815 and 7915, something like this. Very easy to make.
Note: Revision 3. Changed passive xo to 1st order for phase coherence. This is after comparing the female voice with Linkwitz Pluto. The original schematics is still available here, also Rev1, Rev2
I'm too lazy to draw PCB and use a breadboard instead. Don't worry the electrons don't mind... That extra op-amp IC was just for experimentation with notch filter.
Mid & High
I am a follower of simpler-is-better philosophy so driver selection is critical for the Mid & High section. Simple full-ranger like Fostek, Lowther, or Visaton B200 is the obvious choice but they are expensive and hard to find in Australia. So I went with the following:
For midbass use P13WH-00-08 woofer which has very smooth rollof at 6kHz. This driver has been praised by many for it's natural rollof and don't require crossover at all. No crossover is the best thing after active crossovers. Additionally, since it goes up to 6KHz the system will behave as dipole up to that frequency range - which is very good.
Tweeter selection is no longer critical at 6khz. Tweeters usually become expensive when they need to deal with low-frequencies. Since the P13WH midbass can go up to 6kHz the tweeter can use Vifa TC20TD05-06.
As shown on the diagram above it is crossed passively using simple 1st order crossover. Since this tweeter sensivity is 90dB we need to attenuate it using -4db L-pad resistor circuit. Simple.
What can be improved
I have few measurements using ARTA but could not find anything offending the frequency response. Being dipole it's quite hard to nail down true response due to many factors. I simply find music recordings become very enjoyable and I can listen to it for hours and hours.
Try it. It is actually very hard to get things wrong with OB. As the masters say:
"It is difficult to screw up an open baffle speaker design to where it sounds worse than your typical box speaker. " - Sigfried Linkwitz
"Being that open baffles are so easy to design and construct, the crossover is by far the biggest hurdle." - Nelson Pass
But obviously I have my further desires :) ... like:
- Bigger woofers. I wish I had bought 12" ones as there are still some space on the baffle. Or perhaps 2x 12" + 2x 10".
- Perhaps a full-ranger instead of midbass+tweeter. It's difficult to place the tweeter at ear level -- if I have the money that is
- what else... I don't know actually. Perhaps prettier baffles ...
- Stealth 3 ? ... it will be 4x 15" Eminence Alpha with fullrange drivers for mid & high. If I have the room ! perhaps even a CS2 clone.
There is a discussion on this topic at AudioCircle forum.
Bill of Material:
- 4 x 15-inch Eminence Alpha-15A Drivers,
perhaps could be substituted with DJCity B15-100. update: No, the quality is not that good.
- 2 x Selenium D220Ti-8 1" Titanium Horn Driver
- 2 x 12" Round Waveguide 1" Threaded -discontinued?
substitute ? DJCity waveguide
- 48 x 18.5 x 2.5-inch baffle
- Behringer DCX2496, 1 kHz xover, 48 db/octave
substitute with Linkwitz analog filters?
Logistics and costs (Australia)
4x Eminence Alpha 15A : AUD400 - obtained 30/09/08
2x Selenium drivers :
2x 12" waveguides :
Behringer DCX2496 : AUD450.
Cheaper option is to build corresponding analog filters.
After running prototypes and crossing the Alphas to 1kHz with appropriate notch filter I am not convinced that running them that high is beneficial. While tonality is good and I was surprised of their resolution as 15" driver at that frequency, beaming is a real and audible problem. I since then pursue the design where drivers are used at their optimised 1/2 wavelength formula.
Linkwitzlab is not affiliated with Gainphile Blog.
So many good things has been said about this Linkwitz experiment ...
One say it gets too uncomfortably close compared to Orion Dipole ($8,000 system!). I am curious to see how it will compare to my OB speakers.
Importing a pair of Auras is not an issue ($60), but the woofers are too expensive for me. So I substituted them with a pair of Dick Smith woofers laying around. The base uses paving block readily available in the garden. PVC pipes and rubber coupler are available in Bunnings (Australia).
As usual, breadboard implementation. Easy, no soldering. The electrons don't mind.
Halfway build. Very easy, very cheap. $50 worth of PVC from Bunnings.
Update 13/6/2008 - The Verdict
So I've been living with the Plutos for few weeks now. These are my impressions about this amazing little speakers.
- First, it is very small and slender. Visually appealing and cocky. I ended up not painting the PVC pipes as the beige colour is very nice.
- Bass is amazingly deep. But it does not have that dipole speed and transparency. Slam is better than dipole.
- It sounds lively like a dipole. There is no "narrow" box sound.
- However they are very prone to positioning. While initially were intended for small room, I found out that they are less tolerant too, which somewhat defeats the purpose. They ended up in my larger living room where they can "breathe" more freely. I found dipole is much more tolerant even in smaller rooms.
- In summary they sound like something in between dipole and box speaker, with leaning towards dipole.
Schematics posted on request. This is the same schematics available at LinkwitzLab. The greyed ones are not necessary, assuming you have a pair of stereo amps. I also removed the green ones because it sounds better subjectively without those. The green ones are tweeter phase delay and woofer notch filter. I'm not saying I know better than SL, just what I hear in my setup and my room.
I added a shelving low-pass on the woofer circuit to give deep bass. But can't play too loud (it's loud enough for 5x6m room). No need for subwoofer. Use the non-inverting circuit.
(note: removed and replaced with Linkwitz Transform circuit in next update)
Small clip is available at YouTube. Pardon my digicam's limitations and the fact that they only take 30-second clips.
Revisited the equalization, and modified the Linkwitz Transform down to 20Hz. See design and conclusion here. In short, the low-frequency is extended but room modes issues still exist and manifest in boomy room sound.
A friend came over to hear the speakers. He was very impressed, in his own words "I didn't know speakers can image like that." Hence, he is currently building his Pluto2 (real one).
Damping is an interesting topic. This is my final damping fill using polyester fibers. There is also an extensive discussion about the topic here. Note that I am using readily-available Polyfill, not the more exotic acousta-stuff.
The pluto "imposter" was demo'ed at SNA Melbourne get-together (GTG). This is one of the comments:
"Now those little Plutos with some tracks had no right to sound as good as they did.Especially with that vocal track by Angelique Kidjo "Blewu" Amazing what you can make with some dunny pipes!!" -Joz
I am on business assignment for a few months in Switzerland. I brought the Pluto clones and their small footprint and light weight make them a great pair of travelling loudspeakers!
Directivity plot, indoor, normalised at 0 degrees.
The Pluto maintains omni-directional pattern until 4khz where it starts to transition to forward-directional. The transition is excellent with minor widening around 15khz.
Design the attenuation by using the online calculator something like this, using dB attenuation steps of 56-39-26-12-0:
So the theoretical R values from gnd to signal in ohms are:
But for nearest real value use:
This is better than my blue-aps pots. Total cost $4.
Linkwitz' mid test
Lynn Olson' Ariel
North Acoustics Spirit
Troels Gravesen test
Daline Transmission Line
My own S7 OB Speakers
First attempt to use H-Frame was successful but I did not like it aesthetically. So I went back to plain OB.
The final setup is tri-amped system with plain open baffle. Mid & high are actively crossed at 3.5kHz using 24db/oct linkwitz-riley filter. The woofer is crossed at 120Hz and equalised towards low frequency at 6db/octave.
I use a pair of DSE 5" bookshelf speakers and mount it on a 28x45cm baffle. Predictably the bass is lacking but I rather have a natural-sounding midrange than closed-sound of boxed speakers.
It doesn't look too bad as well. Some possible next experimentation to fix the bass are:
- dipole subwoofer
- active compensation circuit
I did try using normal subwoofer but it simply ruins the overall sound character.